The present invention relates to a snapshot creating method, a storage system, and a disk apparatus. In particular, it relates to a snapshot creating method, a disk apparatus, and a storage system in a network-attached storage system for performing a file service via a network.
U.S. Pat. No. 5,790,773, U.S. Pat. No. 5,819,292, and U.S. Pat. No. 5,835,953 have disclosed a computer file system, a snapshot generation, and a mirror disk component. As a prior art on a storage system, there has been proposed and known the network-attached storage (NAS). Based on the NAS, a special-purpose server for providing a file service to a disk apparatus via a network becomes capable of accessing data within the disk apparatus from the network.
Generally speaking, data stored in a storage apparatus necessitates the execution of the back-up processing in preparation for the case where the data has been lost by a malfunction in the apparatus, a defect in a software program, an erroneous operation, or the like. Here, the back-up processing is a processing for copying and saving the data regularly into a magnetic tape or the like so that the lost data will be able to be recovered. The back-up processing is similarly needed in the NAS as well.
If, during the back-up processing task, the data recorded in the storage apparatus has been updated, there occurs an inconsistency in the copied data. Accordingly, it is necessary to prevent the data from being updated during the back-up task so as to guarantee a stored image of the data at a back-up starting point-in-time. In order to avoid the updating of the data to be backed up, it is all right to shutdown all of the programs other than the back-up program which are going to access the data. In the case of a high-availability requested system, however, it is impossible to shutdown the programs for a long time. This situation requires a mechanism for not preventing the programs from updating the data during the back-up, and at the same time for saving the stored image of the data at the back-up starting point-in-time. Usually, a stored image, or a temporal copy, of data at a point-in-time is referred to as “snapshot”, and the data that has been targeted for the snapshot creation is referred to as “operational data”. In the NAS, as methods of providing a state where it is possible to update the data while saving the snapshot at a specified point-in-time, the following methods have been known: Duplexing of the data by the disk apparatus, saving of the pre-updated data by the NAS processing unit, and the like.
Next, the explanation will be given below concerning the snapshot creating methods according to the above-described prior arts.
As the prior art on the duplexing of the data by the disk apparatus, there has been known the technology described in the above-described U.S. Pat. No. 5,790,773. In this prior art, from a normal state before a snapshot creating instruction is issued, all the data within the disk apparatus are duplexed (i.e., mirrored) into a second storage region. Moreover, in this prior art, when creating a snapshot, a management program within the disk apparatus provides one region for the operational data, and another region for the snapshot data.
Meanwhile, the saving of the pre-updated data by the NAS processing unit is performed by saving the pre-updated stored contents into a different storage region at the time of a data updating after a snapshot creating instruction has been issued from the outside. Logically, the snapshot is accessed as different data that is independent of the operational data. However, a part of the snapshot whose operational data has been not updated after the snapshot creating instruction shares its storage region with the operational data. Only a part of the snapshot whose operational data has been updated is created by making reference to the pre-updated stored contents saved into the different storage region. In the case of the NAS, there exists a snapshot technology in the file or directory unit. Namely, when a snapshot creating instruction is issued from the outside, the NAS processing unit, if a file updating occurs thereinafter, saves the pre-updated file image into a different storage region so as to prevent the data crash from occurring.
In this way, the snapshot technology allows the execution of the snapshot creation. In the above-described prior art regarding the snapshot creating method based on the data duplexing processing by the disk apparatus, the duplexing processing must be executed all over the entire physical volumes. Also, a physical volume becomes necessary whose size is the same as that of the physical volume into which the operational data has been stored. This results in a problem that the necessary storage capacity becomes 2 times larger. Also, in this prior art, when the disk apparatus performs a snapshot creation, the disk apparatus cannot differentiate the data contents of the respective blocks within the physical volume. As a result, the data duplexing is executed with respect to the case of a data block as well which is meaningless to the NAS processing unit, the examples of which are: The case where the data contents within the block have been formatted by the NAS processing unit, the case where, although there exists the data within the block, the data has been not used by the NAS processing unit, and the like. This gives rise to a problem of increasing a CPU load in the disk apparatus, a data communications amount within the disk apparatus, and the like.
Also, the above-described prior art regarding the snapshot creating method by the NAS processing unit is carried out under a condition that the NAS processing unit performs all of the saving/management of the pre-updated data. This condition increases a CPU load in the NAS processing unit, and a data communications amount in the network for connecting the NAS processing unit to the disk apparatus. As a result, there exists a problem that the execution speed of an application program is exceedingly lowered in comparison with the one in a snapshot non-creating system. In particular, there occurs a problem that the performance reduction becomes conspicuous in a processing accompanied by a large quantity of data updating, e.g., a replication of the database.